Glossary › world
GLOSSARY
world
appears in 16 paper titles
Definition
Appears mostly in "world model": an internal, learned simulator of how an environment evolves under actions. Having one lets an agent plan by imagining rollouts instead of paying for real trials, which is why the idea recurs in reinforcement learning, robotics, and video generation. A second sense shows up in open-world, meaning the set of situations or categories is not fixed in advance.
Explainers using this term
- Paper Walkthrough: Puffin-World — A World Model That Remembers Which Way Is UpPuffin-World: Scaling a Unified Multimodal Model with Native 3D World States
- SolarWM: Training on 5 Seconds, Walking for an Hour — and Opening the Whole StackSolarWM: Open Data and Scalable Training for Long-Horizon Video World Models
- Paper Walkthrough: ZimaBlue — Turning 120,000 Hours of Egocentric Video into Robot SkillZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training